Report on the technical assessment of the proposed forest reference emission level of Indonesia submitted in 2016
B. Proposed forest reference emission level
II. Data, methodologies and procedures used in the construction of the proposed forest reference emission level How each element in the annex to decision 12/
2. Transparency, completeness, consistency and accuracy of the information used in the construction of the forest reference emission level
Methodological information, including description of data sets, approaches and methods
15. According to the approach applied, forest loss can occur only once in each specific area. The submission covers only natural forests, defined as forests never having undergone unstocking, not even temporarily. During
an effective approach. As a result of the facilitative exchange, Indonesia included a step-by-step description of the methodology applied in the modified submission. The AT considers that the explanations improve the transparency of the submission and commends Indonesia for these efforts.
16. The AT understands that deforestation and forest degradation activity data were quantified by overlaying maps from different time points. In a first step, for each of the time points, a set of satellite images was used to generate land cover maps. In a second step, changed areas were determined from overlaying these maps with others. During the technical assessment, the AT pointed out that direct comparison of satellite images would reduce the sources of errors and is seen as best practice in remote sensing analysis. Indonesia acknowledged that such an approach is often considered to generate better results, but explained that processing data for the complex land-cover information of the whole country and for the long time frame in question would be highly challenging and would require extensive ground truthing. Indonesia highlighted, however, its plans to improve land-use and land-cover mapping. The AT acknowledges this explanation and commends Indonesia for these efforts, agreeing that moving from single-date interpretation to interpretation techniques for time series of satellite images is an area for improvement.
17. During the TA, Indonesia provided a large amount of background information and material, including land-cover matrices for peatlands.
The AT observed some discontinuity in these data that may indicate difficulties in ensuring consistency.14 According to the FREL submission, land-cover maps were derived largely using visual interpretation, with a large number of image interpreters contributing to the task. The AT acknowledged the immense effort that went into building the time series of land-cover maps, but noted that maintaining a consistent interpretation approach across diverse interpreters is very difficult and that this is a potential source of the observed discontinuity. In response, Indonesia explained that it is exploring ways to develop the existing NFMS, for example by combining automated methods and visual interpretation, in order to improve the land-cover data. In line with this, in the view of the
14 For example, the data show that: estate crops increased from 131,000 ha to 493,000 ha in four years, between 1996 and 2000;
plantation forest increased from 281,000 ha to 482,000 ha in two years, between 2009 and 2011; and wet shrubs increased from 0 ha to 479,000 ha and further to 1,401,000 ha in six and ten years, between 1990 and 1996 and between 1990 and 2000.
AT, improving quality management is an area for improvement. The AT commends Indonesia for its efforts to compile land-cover databases using automated interpretation approaches across long time series of satellite images and to improve the existing NFMS. Introducing automated processing may help to ensure a coherent interpretation.
18. Emission factors used in the FREL for deforestation and forest degradation are based on data from Indonesia’s NFI. In the submission, Indonesia reports that 4,450 measurements of permanent sample plots were available for calculations, complemented by additional forest research data. Despite the large number of plots, the AT noted that there are still gaps for individual strata. For example, there are very few measurements for the island of Java or for mangrove forests. The submission highlights ongoing work to improve the NFI and to refine the set of emission factors.
The AT acknowledges these plans and agrees that improving the set of emission factors is an area for improvement.
19. In its submission, Indonesia includes only the gross emissions from deforestation that are associated with clear-cut forests, excluding any potential biomass regrowth and associated removals after the deforestation event. The AT noted that this is a reasonable approach when forest is converted to cropland with annual crops, but not necessarily when the crops are perennial or when there are other woody biomass stocks. In response, Indonesia clarified that when natural forest has been converted to non-forested lands, they rarely grow back into natural forest. Therefore the proposed FREL excludes plantation forests and non-forest classes such as croplands, agricultural lands, shrubs, savannah and grasses, paddy fields, transmigration areas, settlement areas, ports and harbours, mines and bare lands. The AT notes that this exclusion of removals in post-conversion carbon stocks likely leads to an overestimation of emissions from deforestation. The AT considers that it would be useful to assess whether post-conversion removals are significant and could be taken into account when estimating emissions from deforestation, and notes this as an area for technical improvement.15
20. To identify activity data for soil organic carbon emissions from peatland drainage in areas subject to deforestation or forest degradation, the FREL submission overlays land-cover maps with Indonesia’s peatland map. For deforestation and forest degradation occurring on peatlands, additional emissions from the soil organic carbon pool are then calculated. These two data sets are not perfectly harmonized.16 During the TA, Indonesia acknowledged that such imperfect harmonization may occur in “sliver areas”. While maps and definitions could technically be harmonized, Indonesia explained that fully harmonizing data sets would require collaboration and arrangements between several ministry-level agencies and is therefore not easy to achieve. The AT acknowledges that Indonesia has plans to improve peatland mapping and agrees that ensuring consistency between the land-cover map and the peatland map is an area for improvement.
21. Indonesia’s FREL includes soil organic carbon emissions from peatland decomposition associated with deforestation and forest degradation on peatlands. Indonesia calculates these emissions for all areas of deforestation or forest degradation that occur on peatlands using detailed information on the post-conversion land cover, which determines decomposition rates. The decomposition rate is variable; for example, the rate in peatlands with annual crops may differ from that in peatlands with secondary forest. The AT commends Indonesia for its efforts to track land-cover changes post-conversion. The AT noted that while the emission factors from the Wetlands Supplement are intended to be applied only to “drained organic soils”, the FREL does not distinguish areas with and without drainage. During the facilitative exchange for the TA, Indonesia explained that deforestation and forest degradation on peatlands are usually accompanied by drainage. The Party also included information in the modified FREL submission explaining that it is impossible to trace back the drained and the negligibly small areas of undrained secondary peat forest and therefore Indonesia considers it justifiable to consider all of the secondary forests as drained forests. The AT notes that the FREL submission does not currently include data to substantiate this.
The submission does, however, highlight Indonesia’s plans to improve the data on peatlands, chiefly with regard to emission factors and their
16 For example, the calculations identify dryland forests growing on organic soils.
dependency on the water table. The AT acknowledges these plans and agrees that collecting more detailed data on the management of peatlands is an area for improvement.
22. Most of Indonesia’s land-cover information covers time periods of up to six years. The emission factors from the Wetlands Supplement describe annual emissions for land use types. Applying one single emission factor across a multi-year period would introduce bias. Indonesia avoids such bias by averaging the emission factors from before and after the change and then applying the average to the whole period. The AT commends Indonesia for its diligence in avoiding bias.
23. The submission uses different approaches to calculating means for the biomass pools over the reference period. For calculating average deforestation emissions and degradation emissions in the aboveground biomass pool, a simple arithmetic mean is used across all available periods, although the underlying forest-cover change is measured for unequal time intervals according to the availability of land-cover maps.
For example, the estimate for the years 1990–1996 is taken into account only once for aboveground biomass. For calculating peatland emissions in the soil organic carbon pool, however, the applied regression approach reuses the estimates according to the different lengths of time interval.
For example, the estimate for the years 1990–1996 is taken into account six times for the soil organic carbon pool. The AT notes that it would be useful if the submission explained the reasons for such methodological differences.
24. During the TA, Indonesia pointed out that the first biennial update report (BUR) of Indonesia and the FREL use the same principal data sources.
The AT commends Indonesia for establishing overall consistency of data sources between the FREL and the national greenhouse gas (GHG) inventory contained in the BUR.
25. The AT notes that the FREL applies a different forest definition than that used in the national GHG inventory contained in Indonesia’s first
for conversion of forest to other land-cover categories in the GHG inventory.17 The GHG inventory estimates emissions and removals from forest land remaining forest land, taking into account statistical wood harvesting data and adjustments for illegal logging. Whether these estimates are fully consistent with the FREL data for forest degradation cannot be assessed, because the forest degradation estimates for the FREL are prepared using a different methodology that applies emission factors based on different average carbon stock values according to land-cover changes from primary to secondary forest, as discovered by remote sensing. In the modified FREL submission, Indonesia addressed this concern by explaining that emissions from further degradation of secondary forest were not included in the construction of the FREL because, at present, Indonesia does not have the capacity or data to assess the different levels of degradation occurring within secondary forests. The AT commends Indonesia for including an explanation of the forest definition in the modified FREL submission.
26. Indonesia carries out an uncertainty analysis of its FREL. The AT commends Indonesia for the exemplary effort made in providing the uncertainty analysis. During the TA, several aspects of the uncertainty analysis were discussed and areas for future improvement were identified; for example, moving from overall map accuracy to classification accuracy of individual classes or differentiating sampling errors between the individual forest classes. The AT notes that, in line with the stepwise approach, the current uncertainty analysis covers only selected sources of error, and the AT sees broadening the scope of the uncertainty analysis to cover further potential sources of error, and differentiating between land-cover classes with regard to emission factors and activity data, as an area for technical improvement.
Description of relevant policies and plans, as appropriate
27. The FREL submission provides information on legal, policy and planning frameworks related to the forest sector since 1967, and highlights the increasing demand on land and natural resources to meet the population’s food, energy and other development needs. The submission
17 According to the BUR, the deforestation rate fluctuated between 335 and 1,106 thousand ha/year from 2000 to 2012, while in the FREL the deforestation rate fluctuated between 444 and 914 thousand ha/year in the same period.
also includes information on the development of forest planning, management and allocation of lands of national forest for conversion for development purposes. In response to a question raised by the AT during the TA, Indonesia explained that the causes of deforestation and forest degradation have been considered in the construction of the FREL, and highlighted the fact that there was no specific quantitative calculation of deforestation and forest degradation drivers in the submission, because these drivers are very complex, involving cross-cutting issues related to policy, economic growth, population growth and individual sectors, as well as requiring appropriate data sets and research outputs to support them. The AT commends Indonesia for providing information on planning and policies and legal developments that have implications for deforestation.
3. Pools, gases and activities included in the construction of the forest refer-